How to Deploy an MCP Server with Docker

Learn how to deploy an MCP server using Docker, configure Streamable HTTP, build a production-ready container, and verify your deployment. This step-by-step guide uses NitroStack's MCP server implementation and real Docker configuration.

Sharvari Raut
Sharvari Raut
Architecture illustration of deploying an MCP server using Docker and Streamable HTTP.

How to Deploy an MCP Server with Docker

Building an MCP server is only the first step. To make it accessible to AI applications, you need to deploy it in an environment that supports remote connections.

Quick answer: How do you deploy an MCP server?

You can deploy an MCP server by building your application, configuring Streamable HTTP transport, packaging it with Docker, and running it on a server or cloud platform.

With NitroStack, you can create a multi-stage Docker image, configure the server to listen on 0.0.0.0, use port 3000, and run it over HTTP. Your default MCP endpoint will be /mcp.

What is MCP server deployment?

MCP server deployment is the process of running a Model Context Protocol server in an environment where compatible AI applications can connect to its tools and resources.

A local MCP server typically runs on your development machine. A remotely hosted MCP server requires an accessible network endpoint, appropriate transport configuration, and production infrastructure.

Docker provides a convenient way to package and run your MCP server consistently across different environments.

Why use Docker to deploy an MCP server?

Docker packages your application, dependencies, and compiled output into a container image.

For NitroStack, the multi-stage Docker build:

  • Installs project dependencies.
  • Compiles TypeScript into dist/.
  • Builds UI widgets into src/widgets/out, when present.
  • Installs production dependencies in the runtime image.
  • Starts the compiled MCP server.

This makes the same container image usable across local development and supported cloud infrastructure.

How to deploy an MCP server using Docker

For this walkthrough, we'll use NitroStack's Docker deployment configuration.

Step 1: Create your Dockerfile

Create a Dockerfile in your NitroStack project root:

dockerfile
FROM node:20-alpine AS builder
WORKDIR /app

COPY package*.json ./
RUN npm ci

COPY . .
RUN npm run build

FROM node:20-alpine
WORKDIR /app

ENV NODE_ENV=production \
    HOST=0.0.0.0 \
    PORT=3000 \
    MCP_TRANSPORT_TYPE=http

COPY package*.json ./
RUN npm ci --omit=dev

COPY --from=builder /app/dist ./dist
COPY --from=builder /app/src/widgets/out ./src/widgets/out

USER node
EXPOSE 3000

HEALTHCHECK --interval=30s --timeout=3s --start-period=10s \
  CMD wget -qO- http://localhost:3000/mcp/health || exit 1

CMD ["node", "dist/index.js"]

If your project doesn't include a src/widgets directory, remove the corresponding COPY instruction.

Step 2: Configure environment variables

Configure these settings for your Docker deployment:

env
NODE_ENV=production
HOST=0.0.0.0
PORT=3000
MCP_TRANSPORT_TYPE=http

HOST=0.0.0.0 allows connections from outside the container. MCP_TRANSPORT_TYPE=http configures NitroStack to use HTTP-only transport.

Keep API keys and database credentials outside your Dockerfile.

Step 3: Build the Docker image

Run:

bash
docker build -t my-mcp-server .

This compiles your application and creates the container image.

Step 4: Run the MCP server

Start the container:

bash
docker run -d \
  --name mcp-server \
  -p 3000:3000 \
  my-mcp-server

Your MCP server will be accessible locally at:

text
http://localhost:3000/mcp

Step 5: Verify your deployment

NitroStack provides a health endpoint:

bash
curl http://localhost:3000/mcp/health

A successful response looks similar to:

json
{
  "status": "ok",
  "transport": "streamable-http",
  "sessions": 0,
  "uptime": 4.2
}

You can inspect container logs using:

bash
docker logs mcp-server

Finally, connect an MCP-compatible client and test tool discovery and execution.

How do you deploy an MCP server with Docker Compose?

If your MCP server needs a database, Docker Compose lets you run both services together.

Create a docker-compose.yml file:

yaml
services:
  mcp-server:
    build: .
    ports:
      - "3000:3000"
    env_file: .env.production
    depends_on:
      - database
    restart: unless-stopped

  database:
    image: postgres:16-alpine
    environment:
      POSTGRES_DB: myapp
      POSTGRES_USER: app
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
    volumes:
      - db-data:/var/lib/postgresql/data

volumes:
  db-data:

Run:

bash
docker compose up -d

The MCP server can access PostgreSQL using database as its hostname within the Compose network.

For production, configure database readiness checks, backups, and secure secret management.

Common MCP server deployment problems

Connection refused: Check that HOST=0.0.0.0 is configured and the Docker port mapping is correct.

Widgets not rendering: Confirm that the widget build completes and src/widgets/out is copied into the runtime image.

Container exits unexpectedly: Run docker logs mcp-server to identify startup errors.

Health check fails: Verify that the application is running and /mcp/health is accessible.

MCP server deployment: Quick reference

ConfigurationNitroStack default
TransportStreamable HTTP
MCP endpoint/mcp
Default port3000
Container host0.0.0.0
Health endpoint/mcp/health
Build commanddocker build -t my-mcp-server .
Run commanddocker run -d --name mcp-server -p 3000:3000 my-mcp-server

Frequently asked questions

How do you deploy an MCP server?

Build your MCP application, configure a supported transport such as Streamable HTTP, package it with Docker, and run it on a server or cloud platform.

Can you deploy an MCP server using Docker?

Yes. Docker packages the MCP application and its dependencies into a portable container image that can run on compatible infrastructure.

How do I host an MCP server?

You can host an MCP server on your own infrastructure or a cloud platform that supports your application. For remote connections, configure the appropriate HTTP transport, network access, and security.

What port does NitroStack use for Docker deployment?

NitroStack uses port 3000 by default. You can change it using the PORT environment variable.

What is the NitroStack MCP endpoint?

The default Streamable HTTP endpoint is /mcp.

How do I check whether my MCP server is running?

Use NitroStack's /mcp/health endpoint and inspect your Docker container logs.

Can NitroStack run with PostgreSQL in Docker?

Yes. You can run NitroStack and PostgreSQL as separate services using Docker Compose.

Deploy your MCP server with NitroStack

Docker provides a consistent way to package and run MCP servers. NitroStack makes this workflow practical with its SDK, production configuration, Streamable HTTP support, and built-in health endpoint.

Ready to deploy?

Build your MCP server with NitroStack and take it from development to production.

Tags

Sharvari Raut

Sharvari Raut

Author

A blog poet ✍🏻

Popular tags